neuroimaging of cocaine craving states ......110 neuroimaging of cocaine craving states: cessation,...

110

NEUROIMAGING OF COCAINECRAVING STATES: CESSATION,

STIMULANT ADMINISTRATION, ANDDRUG CUE PARADIGMS

A. R. CHILDRESSTERESA R. FRANKLIN

J. LISTERUDPAUL D. ACTON

CHARLES P. O’BRIEN

Cocaine craving is a cardinal feature of addiction to thedrug and is clinically significant because of its potential linkto relapse. Addicted patients describe the craving as a power-ful, ‘‘must-have’’ pull that causes them to risk, and some-times lose, their relationships, families, money, possessions,jobs, and even their lives. ‘‘Anticraving’’ medications aregreatly needed, but despite intensive research efforts sincethe mid-1980s, such agents have remained elusive (1,2).This dilemma is in part a consequence of our inability todefine or measure cocaine ‘‘craving’’ clearly. Diversity inmeasurement may well account for some of the variabilityin the data collected, as described below. In part, the prob-lem is our nascent understanding of which brain substrate(s)an ‘‘anticraving’’ medication should address. Until recently,the activity of the brain during cocaine craving was a matterof inference rather than direct observations. The increasedavailability of powerful tools for brain imaging in vivo hasthrust research on drug craving forward into a new era.Several laboratories have begun to measure brain activityduring cocaine-craving states directly. This chapter reviewscurrent findings, offering a framework for the results anda discussion of their theoretic and treatment implications.

IS THERE MORE THAN ONE KIND OFCOCAINE CRAVING?

Despite sophisticated tools for brain imaging, the study ofcocaine craving is complicated by the fact that desire for

A. R. Childress, Teresa R. Franklin, J. Listerud, Paul D. Acton, andCharles P. O’Brien: Addiction Treatment Research Center, Departmentof Psychiatry, University of Pennsylvania School of Medicine, Philadelphia,Pennsylvania.

the drug can be reported under dramatically different condi-tions, and measurement is carried out with a range of differ-ent scales. For example, craving may be reported in associa-tion with cocaine cessation (3), but also in association withcocaine administration (4) and cues signaling cocaine (5,6).Brain substrates may differ across these conditions. Interest-ingly, changes in the mesolimbic dopamine (DA) systemsof the brain have been implicated in all three conditions,but the direction of change is not the same in all three. Apossible (tonic) decrease in mesolimbic (nucleus accum-bens) DA has been proposed in the case of cessation/func-tional depletion (7–9), whereas an (episodic) increase in DAhas been hypothesized in the case of cocaine administration(10) and response to cocaine cues (11,12). Of course, thesediffering DA-related states are not mutually exclusive andmay interact in important ways; for example, a dose of co-caine or a cocaine-related cue may have a different brainimpact in early versus late cessation, depending on the dy-namic re-regulation of the affected substrates. The possible‘‘duality’’ (toomuchDA vs. too little DA) of cocaine cravinghas implications not only for the design of imaging studiesbut also for the treatment of cocaine craving (2). If the samebrain substrates are responsible for the craving associatedwith DA deficiency and the craving associated with DAexcess, how can either of these craving states be treatedwithout worsening the other? We will return to this ques-tion after a review of the neuroimaging evidence for cravingacross the conditions.

Neurotransmitter systems other than DA are likely to beinvolved in cocaine reward and motivation (13,14). Forexample, serotonin (15,16), glutamate (17), corticosteroids(18–20), and opioids (21,22) have each received substantialcocaine-related research attention. However, in the neu-

Neuropsychopharmacology: The Fifth Generation of Progress1576

roimaging of cocaine craving, many of the studies have useda DA-related framework for their hypotheses, interpreta-tions, or both. The emphasis on a dopaminergic systemreflects not only a long literature on DA involvement instimulant reward andmotivation (23–25) but also the avail-ability of DA-related neuroimaging tools for humans (26).Imaging studies testing the role of other neurotransmittersand neuromodulators in cocaine craving will be a welcomeaddition to the field. The only currently available study inthis category is included in the current review (27).

The neuroimaging studies of cocaine craving discussedbelow are categorized according to whether their primaryfocus is cessation, stimulant administration, or drug cue para-digms. In studies in which the paradigms may overlap (e.g.,patients in early cessation viewing cocaine cues), this isnoted, as it may be helpful in the final integration of cravingresults across studies and paradigms.

IMAGING OF CRAVING DURING COCAINECESSATION

Hypotheses

Early in the cocaine epidemic, investigators hypothesizedthat the long-term use of cocaine resulted in a functionaldepletion of brain DA (9) in regions critical for the regula-tion of mood, motivation, thought, and concentration.Acute cocaine administration to laboratory animals or tohumans elevates levels of synaptic DA (and other mono-amines) by reuptake blockade (28,29). However, brain mea-surements in long-term users (or laboratory animals main-tained on cocaine) have revealed differences from controls,such as reduced DA synthesis (30), reduced cocaine uptake(31), down-regulation of postsynaptic DA sites (32,33), andaltered responses in the endogenous opioid system (34–36),differences that may reflect the homeostatic attempts of thebrain to cope with a cocaine-induced flood of synaptic DA.DA transporters (DATs) in the long-term user of cocainemay also be dysregulated, although the direction of observedchange varies across studies (35,37–40), potentially reflect-ing the oscillating nature of re-regulation (see also refs.31,41). The DA dysregulation observed in cocaine usersfollowing cessation is supported by numerous animal stud-ies. These studies have found alterations in accumbens DAlevels (7), the threshold for rewarding brain self-stimulation(42), the metabolism of reward-relevant regions (43), DAsynthesis (44), and postsynaptic DA receptors (44). Thus,on cocaine cessation, the DA systems of some cocaine usersmay be in a neurologically adapted, dysregulated state. Thelowered mood, energy, and concentration experienced bysome cocaine patients during abstinence may be related, atleast in part, to a dysfunction in brain DA systems. Accord-ing to this general view, the craving that arises during absti-nence may also reflect DA system dysregulation (8). Doneuroimaging studies support this hypothesis?

Data

The results of several neuroimaging studies are consistentwith the notion of DA dysregulation in cocaine cessation;we review here the studies that have also included cravingmeasures.

Early Cessation (1 Week or Less post Cocaine)

Volkow et al. (45) found a higher rate of metabolism bypositron emission tomography (PET) with 18F-fluorodeox-yglucose in cocaine users in early cessation (�1 week postcocaine) than in controls in twomajor regions of DA projec-tion, the orbitofrontal cortex and the basal ganglia. In thisstudy, craving (‘‘none,’’ ‘‘mild,’’ ‘‘moderate,’’ or ‘‘severe’’)during the week before the scan was positively correlatedwith (both relative and absolute) metabolic rates in the or-bitofrontal cortex, and with (absolute but not relative) meta-bolic rates in another major mesocortical DA projectionregion, the prefrontal cortex.

Malison et al. (39) studied DATs with single-photonemission computed tomography (SPECT) and B-CIT, atracer that binds to dopamine and serotonin transporters,in patients abstinent from cocaine for 96 hours or less todetermine whether they exhibited the transporter elevationspredicted by a neuroadaptive hypothesis. DAT increases ofapproximately 20% in comparison with controls were in-deed detectable, although they were more modest than theDAT increases found in postmortem studies of cocaine users(37,38). An inverse relationship was noted between DATlevel and depression scores, but no relationship of DATlevel to craving scores (on the Cocaine Craving Scale) wasfound (3).

Early and Later Cessation

Long-term cocaine use can alter �-opioid binding (34–36),which indicates an interaction of DA and endogenous opi-oid systems. Given the role of endogenous opioids in moodmodulation, up-regulated �-opioid receptors may contrib-ute to dysphoria or craving in cocaine cessation. Zubieta etal. (27) used PET and 11C-carfentanil (a high-affinity �-opioid agonist tracer) to image �-opioid-receptor bindingin cocaine patients at 1 to 4 days, and then at 4 weeks postcocaine. In comparison with controls, the cocaine patientsshowed up-regulation of �-opioid receptors in the caudate,thalamus, anterior cingulate, frontal, and temporal brainregions; these changes persisted for 4 weeks in all but thetemporal region. On the early cessation scan, craving (Min-nesota Craving Scale on the prior evening, 100-mm visualanalogue scale just before the PET, or both) was positivelycorrelated with �-opioid binding in the amygdala and theanterior cingulate, frontal, and temporal cortex. These fourregions all receive significant DA projections, consistentwith DA–opioid interactions. Scores on the Beck Depres-

Chapter 110: Neuroimaging of Cocaine Craving States 1577

sion Inventory did not correlate with �-opioid-receptoravailability, which indicates that the early cessation cravingwas not simply a depressed mood. None of the correlationsof craving with �-opioid binding were significant for thelater, 4-week scan.

Later Cessation

In abstinence extending beyond a week (imaging at 1 to 6weeks, and again at 3 to 4 months post cocaine), neuroimag-ing studies from the laboratory of Volkow et al. (46) haveshown that frontal metabolism is decreased in the brains ofcocaine users in comparison with that in controls. Althoughcraving (averaging 3 � 1 on a scale of 0 to 7) was assessedas a subject characteristic in this study, no correlations withmetabolic rates were reported. In a further study of thesame patient sample, reductions in orbitofrontal cortex andcingulate metabolism were particularly profound, and thesereductions were correlated with reductions in (striatal) DAD2-receptor availability (33). However, craving (on a scaleof 0 to 10) during the week of the study did not correlatewith striatal D2-receptor availability; correlations of cravingwith metabolic rates were not reported.

Paralleling the metabolic findings of Volkow, Childresset al. found the resting regional cerebral blood flow (rCBF)to be significantly lower in the anterior cingulate (47) andleft medial orbitofrontal cortex (gyrus rectus) (48) of cocainepatients at an average of 13.5 days after cessation in compar-ison with controls. However, neither baseline cocaine crav-ing nor withdrawal (self-rated on a scale of 0 to 9) at thetime of the scan correlated with rCBF in these regions.

Summary of Cessation Studies

The neuroimaging data clearly show a number of differencesbetween the brains of cocaine patients undergoing cessationin comparison with controls not using drugs. The observeddifferences often are clearly linked to brain DA systems.However, the relationship of these brain indices to ‘‘cessa-tion craving’’ has been variable. At this time, correlativeevidence from studies in early (�1 week) cessation, but notfrom those conducted at longer intervals post drug, indicatesa relationship between craving and brain responses (orbito-frontal and prefrontal cortex metabolism, �-opioidbinding).

Because the ‘‘early’’ and ‘‘late’’ cessation studies wereoften conducted in separate populations, and with impor-tant differences across studies (e.g., monitoring of absti-nence, inpatient vs. outpatient population), any conclusionsmust be drawn with caution. Powerful within-subject de-signs, including frequent scans and subjective measuresacross a period of prolonged abstinence, would be helpfulin clarifying cessation–brain–craving relationships. If suchlongitudinal studies were to confirm a lack of a relationshipbetween craving and later resting hypoactivity (by metabo-

lism and rCBF) in DA projection areas, such findings wouldhave implications for both theory and treatment. For exam-ple, if low DA tone is not associated with craving, thenenhancing DA activity might not help craving, and couldeven be problematic (49).

Even if unrelated to craving, the differences between co-caine patients and controls that are evident at both earlyand later time points (e.g., in DA D2 and �-opioid recep-tors) may still be very important. They may reflect otherfunctional consequences of cocaine use, or group differencesthat possibly predate cocaine use or even predispose subjectsto such use. The latter possibility has received very recentsupport. In a study of Volkow et al. (50), initial liking forintravenous methylphenidate (a stimulant also acting by in-hibition of reuptake of DA) in normal persons was inverselyrelated to D2-receptor availability. The D2-receptor levelsfor normal subjects who liked methylphenidate were signifi-cantly lower than those for normal subjects who did notlike the drug, and they were strikingly similar to those oflong-term cocaine users in earlier studies by the same inves-tigators (51). Reduced D2 receptors may thus be a markerfor vulnerability to stimulant misuse in addition to, or per-haps even instead of, a possible consequence of misuse.

IMAGING OF CRAVING DURINGSTIMULANT ADMINISTRATION

Hypotheses

Administering cocaine in the laboratory is a reliable androbust trigger of cocaine desire (4), and cocaine users com-plain that the first dose (‘‘taste’’) of cocaine in a binge rap-idly elicits profound craving, drug-seeking behavior, and asecond dose. Almost two decades ago, Eikelboom and Stew-art (52) modeled this behavior in rats, showing that smalldrug ‘‘primes’’ could motivate drug seeking and reinstateextinguished responding for drug. According to the priminghypothesis, the initial drug effect always precedes the fulldrug effect and comes (through classic conditioning) to trig-ger a druglike brain state. The state has powerful positiveincentive properties, ‘‘pulling’’ the organism back to thedrug.

Koob (53) has proposed a different way in which the firstdose of cocaine might lead to the next. In this ‘‘opponentprocess’’ view, the brain responds to cocaine with homeo-static processes, some of which are the hedonic opposite ofthe direct effect of the drug, euphoria. The opponent re-sponse (i.e., dysphoria) emerges as the direct effects of thedrug dissipate and motivates drug seeking to reduce discom-fort. Clinically, patients do complain about the jittery offsetof the cocaine high, and they recognize that taking anotherdose of the drug will alleviate this state.

So, is the craving associated with human cocaine admin-istration more closely related to a brain state that occurs atthe onset or at the offset of the drug response? Although thequestion is posed as a choice, these possibilities (unfortu-


nately for the task of developing medications) are not mu-tually exclusive. Craving of the positive, appetitive,‘‘primed’’ variety may map onto brain responses associatedwith the initial effect of the drug and be followed shortlythereafter by the dysphoric craving of offset, which maymap onto a later set of brain responses that are opposite indirection to those of drug. The exquisite temporal sensitivityof functional magnetic resonance imaging (fMRI) allowsthese possibilities to be examined directly, as discussedbelow.

What are the likely neuroanatomic and neurochemicalfeatures of the craving state(s) associated with cocaine ad-ministration? More than two decades of animal research(see refs. 25,54–56 for review) support the involvement ofthe mesolimbic and mesocortical DA systems of the brainin cocaine reinforcement and motivation. Thus, a priorineuroanatomic predictions include the familiar projectionsof the DA cells in the ventral tegmental area of the midbrainto the ventral striatum (nucleus accumbens), amygdala,basal forebrain, orbitofrontal cortex, and medial prefrontal/anterior cingulate cortex. And, although recent animal (13)and human (57) research leaves room for the contributionof other brain systems, DA neuronal elements (DATs, post-synaptic D2 receptors) have also been initial neuroimagingtargets.

Data

Several laboratories have now imaged cocaine users duringstimulant administration, but some of these studies eitherdid not obtain a craving measure (58–61), did not analyzethe craving item (62), or analyzed a craving item but didnot attempt to relate it to the brain measure under study(63). The remaining four studies discussed below have beenpublished since 1997.

Breiter et al. (64) used temporally sensitive fMRI tech-nology with a BOLD (blood oxygen level-dependent) scanto map the brain circuitry activated during a period 5 min-utes before, and 13 minutes after, cocaine (0.6 mg/kg, maxi-mum dose of 40 mg) versus saline infusion in cocaine sub-jects abstinent a minimum of 18 hours. Subjective ratings(‘‘rush,’’ ‘‘high,’’ ‘‘craving,’’ and ‘‘low’’) were taken eachminute throughout the experiment. ‘‘Rush’’ and ‘‘craving’’(defined as wanting more cocaine) ratings were later corre-lated with the group-averaged temporal pattern of signalsfrom each brain region meeting specified threshold and ex-tent criteria for differential activation by cocaine. ‘‘Craving’’ratings increased steadily from the onset of the cocaine infu-sion, peaking at approximately 12 minutes post cocaine.No activity in any single brain region precisely echoed theonset and late peak of ‘‘craving’’ ratings. However, signifi-cant positive correlations were obtained with regions havingearly-onset (during euphoria) but sustained activations.These included the nucleus accumbens/subcallosal cortexand some paralimbic sites (a section of the parahippocampalgyrus, a section of the posterior insula, and a section of the

anterior cingulate). Signal change in the amygdala duringcocaine administration was initially reported as heterogene-ous (some patients showed increases and others showed de-creases), not correlated with rush, and negatively correlatedwith craving ratings. However, in a follow-up study withcardiac gating of the fMRI signal (see below), the directionof signal change in amygdala was positive for all subjects.

In contrast to ‘‘craving’’ ratings, ‘‘rush’’ ratings peakedquickly (at 3 minutes) after infusion and then declined rap-idly. Although the brain regions that correlated with ‘‘crav-ing’’ and ‘‘rush’’ overlapped substantially, a clear dissocia-tion was also noted. ‘‘Rush,’’ but not ‘‘craving,’’ correlatedwith early maximal, short-duration signals from the ventraltegmental area and basal forebrain (and sections of the cin-gulate). On the other hand, ‘‘craving,’’ but not ‘‘rush,’’ cor-related with an early-onset but sustained signal from thenucleus accumbens/subcallosal cortex. All the activated re-gions showed early onset to cocaine. Thus, the primary dif-ference between ‘‘craving’’ and ‘‘rush’’ (euphoria) substrateswas not a matter of which regions were activated, but ofhow long. Put another way, a full orchestra is playing fromthe outset of cocaine administration. As the ‘‘rush’’ wanes,some instruments drop out. ‘‘Craving’’ corresponds to thestrains of those that play on.

How do these data fit with the ‘‘priming’’ and ‘‘opponentprocess’’ hypotheses of craving in response to cocaine? Un-fortunately, the fit is not completely straightforward foreither view. ‘‘Craving’’ mapped onto the activity of struc-tures (nucleus accumbens/subcallosal cortex) with an earlybut sustained signal. At the first level of examination, thisfinding seems consistent with a priming effect; the signaloccurs very early and therefore looks like a direct drug effect.However, according to the priming hypothesis, ‘‘craving’’should map better onto the brain correlates (ventral tegmen-tal area and basal forebrain) for the direct effects of ‘‘rush’’and euphoria. In terms of clear evidence for a simple oppo-nent process view, no later-occurring activations opposite tothe direction of the ‘‘direct’’ drug effects in ventral tegmen-tal area/basal forebrain, nucleus accumbens/subcallosal cor-tex, or other brain regions were identified. However, be-cause the direct drug effect was a positive signal change invirtually all brain areas, detecting opposite direction effectswould necessitate the detection and interpretation of fMRIsignal decreases (‘‘negative signal change’’). Because of thephysiologic basis of the BOLD signal, the meaning of ‘‘neg-ative signal change’’ is an ongoing research challenge forfMRI.* Until this issue is resolved, PET in combination

* fMRI is extremely vulnerable to movement artifact. Signals from theamygdala and other structures near the base of the brain can be affectedby the slight movement, at each heartbeat, of blood entering the brainthrough large vessels, and unreliable or uninterpretable data can result.Cardiac gating of the fMRI signal allows the fMRI scanner to be controlledby the heartbeat, and images are collected in the intervals between beats.Under these controlled conditions, the direction of change in the amyg-dala was positive for all subjects (H. C. Breiter, personal communication).


with a 15O bolus performed during and after cocaine admin-istration offers sufficient temporal resolution (15O has ahalf-life of 128 seconds) that it can be used to sort out‘‘early/direct’’ from any ‘‘later/opposed’’ effects of cocaine.Another possible complication in detecting ‘‘opposed’’ drugeffects is that opponent processes can be conditioned; dur-ing the course of thousands of cocaine administrations, aresponse may ‘‘move back in time’’ so that it becomes nearlyengaged at drug onset and persists. In experienced users (theonly subjects who can be given cocaine in human studies),distinguishing ‘‘direct’’ from ‘‘opposed’’ effects could bevery difficult. Animal research mapping the temporal corre-lates of brain response to cocaine and its signals during thecourse of initial and repeated administrations could clarifythese relationships; of course, such experiments cannot ethi-cally be conducted in humans.

The other three neuroimaging experiments involvingstimulant administration to cocaine users and measures ofcraving have been conducted by the Volkow team at Brook-haven Laboratories. In one of these experiments, cocainewas used as the stimulant probe; in the other two, methyl-phenidate was used. In the cocaine study, cocaine users weregiven intravenous cocaine in doses of 0.3 to 0.6 mg/kg (adose range known to induce euphoria), administered to-gether with a tracer dose of 11C-cocaine to measure DAToccupancy (29). Subjective self-ratings were taken everyminute for the first 20 minutes post infusion, and then at10-minute intervals for the next 40 minutes. The ratingsof cocaine ‘‘high’’ and ‘‘rush’’ were positively correlated withDAT occupancy, but ‘‘craving’’ (the desire for cocaine, ratedon a scale from 1 to 10) and ‘‘restlessness’’ were not. Thus,as in the earlier study of Breiter et al. (64), ‘‘craving’’ didnot map onto precisely the same substrate as ‘‘rush’’ and‘‘high.’’ This variability among studies may be a conse-quence of differences in the way craving was measured ineach study.

The results from studies of methylphenidate administra-tion suggest that a simple DA hypothesis of stimulant effects(including craving) may be insufficient. One study com-pared the subjective and brain DA response of cocaine users(3 to 6 weeks post cocaine) with that of controls after aninjection of methylphenidate, which (like cocaine) blocksDA reuptake (51). An intravenous injection of 0.5 mg ofmethylphenidate per kilogram was followed by an injectionof 11C-raclopride, a dopamine D2 ligand sensitive to com-petition by endogenous DA. Regions of interest were thestriatum, thalamus, and cerebellum (as a comparison regiondevoid of D2 receptors). Subjective measures were taken5 minutes before and 27 minutes after methylphenidate.‘‘Craving’’ in response to methylphenidate was muchgreater in the cocaine users than the controls and was corre-lated with an enhanced response (reduced raclopride bind-ing) in the thalamus. ‘‘High’’ did not correlate with eitherthalamic or striatal raclopride binding, but a possible corre-lation may have been compromised by taking the subjectivemeasures at 27 minutes, when the high had waned. Interest-

ingly, both the ‘‘high’’ and the DA response tomethylpheni-date (measured by raclopride binding) in the striatum weregreater in the controls than in the cocaine users, as thoughthe cocaine users’ response to methylphenidate had beenblunted.

In a second study of cocaine users (abstinent on averagefor 14 days), two sequential doses (0.5 and 0.25 mg/kg) ofmethylphenidate were administered 90 minutes apart, afterwhich metabolism (measured by PET and 18F-fluorodeox-yglucose), D2-receptor availability (measured by 11C-raclo-pride), and subjective responses (27 minutes after each infu-sion) were determined (66). The actions of methylphenidateon brain metabolism showed substantial variability acrosssubjects that correlated with striatal D2-receptor availabil-ity; the stimulant increased metabolism in subjects with ahigh level of D2-receptor availability and reduced it in sub-jects with a low level of D2-receptor availability. Althoughmethylphenidate induced metabolic increases in severalareas (cingulate, thalamus, cerebellum), it increased rightorbitofrontal and right striatal metabolism only in the sub-jects who experienced cocaine craving (‘‘desire for cocaineand perception of loss of control over cocaine’’). This obser-vation indicates a possible (lateralized) role for these regionsin stimulant-induced craving, but it also shows that an in-crease in DA secondary to reuptake blockade is not in itselfsufficient to induce the metabolic increases in the frontalregions. As in the prior study, ratings of ‘‘high’’ at 27 min-utes post methylphenidate did not correlate with the effectsof the drug on metabolism or D2-receptor availability.

Summary

The only fMRI study performed during cocaine administra-tion showed widespread brain activation, including activa-tion of the classic mesolimbic–mesocortical circuitry oftenimplicated in the reinforcing and motivational effects ofcocaine. Although several brain regions were commonly ac-tivated by both craving and rush, rush-associated ventraltegmental area/basal forebrain signals rapidly declined whilethe craving-associated signals in the nucleus accumbens/subcallosal cortex persisted. This partial dissociation sug-gests at least some independence of substrates for these twostates. The early onset of brain activation in cocaine-inducedcraving is consistent with the priming hypothesis; the partialdissociation with rush/euphoria is not. The lack of apparent‘‘offset’’ activations correlated with craving is inconsistentwith a simple opponent process view, but conditioned op-ponent effects could occur very early in highly experiencedusers, so that their origins would be obscured.

Studies that use a priming dose of cocaine necessarilyimage both the direct effects of the drug on the systembeing measured (e.g., DA system) and simultaneously anycognitive–affective–craving brain activity related to the co-caine cue. This dual time course of effects may account forsome of the variability or inconsistency in effects reported.

With regard to which dopaminergic neuronal elements


are involved in the craving response, the results are mixed.Cocaine-induced craving was unrelated to DAT occupancy.However, methylphenidate-induced cocaine craving wascorrelated with right striatal and right orbitofrontal in-creases in metabolism and with an enhanced response (in-dexed by 11C-raclopride) in the thalamus in comparisonwith controls. These findings could support a postulatedDA dysfunction in the striatal–thalamic–orbitofrontal cir-cuit in cocaine addiction (33,67). One similarity betweenthe craving findings in early cocaine cessation (reviewedearlier) and the craving results in stimulant administrationis a positive correlation with orbitofrontal activation. Asdiscussed in the final section, the orbitofrontal cortex is thusfar the only region linked to craving in all three paradigms:(early) cessation, stimulant administration, and exposure todrug-related cues.

Based on the neuroimaging data during stimulant ad-ministration, a role of transmitters other than DA in meth-ylphenidate-induced effects (both high and craving) seemslikely. Methylphenidate alone was insufficient to increasefrontal metabolism, and in other studies by the Brookhaventeam, significant levels of DAT occupancy by intravenousmethylphenidate did not always result in a subjective high inhealthy controls (craving was not probed) (68). The studiesneeded to determine whether these variable effects are alsolikely for cocaine (namely, giving pharmacologic doses ofcocaine to healthy controls) cannot be performed; primatestudies offer an important alternative.

IMAGING OF CRAVING DURING COCAINECUES

Hypotheses

A guiding hypothesis of much of the research in cue reactiv-ity is that the powerful craving and arousal responses todrug-related cues are based on simple classic conditioning(5,6). According to this view, cocaine cues trigger cocaine-related subjective and physiologic responses, including crav-ing, because they reliably predict the arrival of the directeffects of the drug. As both the animal and human literaturehas shown, however, ‘‘simple’’ classic conditioning is oftenanything but simple. Drug-related conditioning can resultin both responses similar to those produced by the drugitself (‘‘druglike’’) and responses opposite to those of thedrug (‘‘drug-opposite’’) (69), likely reflecting a conditionedcompensatory response to drug onslaught. Both kinds ofresponses may be of motivational significance. Druglike re-sponses to cues reminiscent of the drug (‘‘Wow . . . It’s likeI’m feeling it already . . . and I haven’t even had any yet! Ican’t wait!!’’) may act as a powerful positive incentive, pull-ing the user back to the drug. If drug-opposite responsesto the cues are uncomfortable (in opiate users, these includetearing, yawning, sweating, and nausea), they may alsoprompt drug seeking.

Is cue-elicited cocaine craving more ‘‘druglike’’ or ‘‘drug-opposite’’? Do the multiple cues surrounding the cocaineexperience become linked predominantly to the brief, in-tense, orgasmic euphoria . . . or to the jittery, sometimesuncomfortable, offset effects of this very short-acting drug?Both links seem possible, but they would yield oppositepredictions for neuroimaging and pharmacotherapies. Ifcraving to common external (paraphernalia, other users,drug-buying locations, drug talk) or internal (e.g., drugdreams, recurrent memories of ‘‘high’’) cues is predomi-nantly druglike, we would expect that some elements of themesolimbic–mesocortical DA system activated by cocaineitself are also activated during cue-induced craving. In ani-mals, cues for cocaine can indeed trigger mesolimbic DAoverflow in nucleus accumbens and amygdala (70) and canactivate c-fos (an immediate early gene) in the cingulate (71,72), amygdala (72), and nucleus accumbens (71). Patientsoften describe cocaine-like effects in response to cues, in-cluding heart pounding, ear ringing, head buzzing, stomach‘‘flipping,’’ mild euphoria, a ‘‘taste’’ of cocaine in the backof the throat, even the ‘‘smell’’ of cocaine in the room . . .and, of course, profound desire. But what do the brains ofcocaine users say about the nature of cue-induced cocainecraving?

Data

As shown in Table 110.1, the first neuroimaging study inwhich drug cues were used to induce craving was presentedin 1992 (73,74). It tested whether video-induced cocainecraving might increase endogenous DA, as indexed by com-petition with a D2 ligand, 123I-iodobenzamide, in SPECT.The low signal-to-noise ratio of 123I-iodobenzamide, thelow resolution of SPECT, and timing of the cue activation(after uptake of the tracer) likely undermined the ability ofthis early study to detect increased DA release in responseto cocaine cues. After this initial effort, imaging studies ad-dressed the neuroanatomic rather than the neurochemicalsubstrates of cue-induced craving.

At least six additional laboratories (Childress et al., Vol-kow et al., Kilts et al., Grant et al., Garavan et al., Wanget al., Maas et al.) have conducted imaging studies withcocaine cues since 1992. These studies cover a range ofimaging technologies (PET with 15O bolus, PET with 18F-fluorodeoxyglucose, fMRI) and include several variationson the method of presenting drug cues to induce craving.The variations are useful because results obtained in onlyone laboratory, or with one method of cue presentation,may be more related to the specifics of the laboratory orstimuli than to the psychological state of interest (cue-induced cocaine craving). Conversely, any replication andconvergence of findings across multiple laboratories andmethods are very encouraging.

A survey of the data in Table 110.1 reveals several conver-gent findings and suggests that brain activation in response


TABLE 110.1. BRAIN IMAGING OF CUE-INDUCED COCAINE CRAVING

Imaging Cocaine Days ofLaboratory Technology Population Cessation Cue Description Resultsa

U. PennsylvaniaChildress et al., SPECT, Treatment- Range, 6–51 15’ personalized audio 123I-IBZM not

1992 (73,74) 123I-IBZM seeking followed by 10” video displaced fromcompetition (n = 10; (separate days for striatum

10 controls) drug/nondrug)Childress et al., PET, 15O bolus Treatment- Average, 13.5 25’ narrative nondrug Limbic: amygdala +

1994–1999 (ROI) seeking video, then 25’ cocaine a. cingulate +(47,79) (n = 14; video (both with OFC 0

6 controls) soundtrack) hippocampus 0Comparison: striatum—

DLPFC 0cerebellum 0visual cortex 0

Childress et al., PET, 15O bolus Treatment- Range, 7–22 (as above) GABAB agonist1999–2000 (ROI) seeking baclofen may(103) (n = 7) blunt limbic

activationListerud et al., fMRI, BOLD Treatment- Average, 14.5 15’ nondrug video, amygdala +

2000 (80) seeking Range, 3–38 then 15’ cocaine a. cingulate +(n = 7; video12 controls)

NIDA ARCGrant et al., PET, FDG Nontreatment 36–48 h 10+ repetitions of a DLPFC +, ^

1996 (83) (ROI) (n = 13; (verified) 2.5’ silent video VLPFC +5 controls) plus paraphernalia; m. OFC +

snort option m. temporal (separate days for (amygdala) + ^drug/nondrug cues) retrosplenial +

temporal/parietal +temporal +extrastriate/striate +peristriate +cerebellum ^r.a. OFC ^(–)

Harvard–McLeanMaas et al., fMRI, BOLD Nontreatment Alternating 2.5” drug/ a. cingulate +, ^

1998 (86) (ROI) (n = 6; nondrug video clips DLPFC +, ^6 controls) from Childress tapes cerebellum 0

(faces blurred)Emory University

Kilts et al., PET, 15O bolus Treatment- Range, 7–17 1’ guided imagery (by r. amygdala +1996–2000 (82) (SPM) seeking audiotapes of nondrug, l. a. cingulate +

(n = 8) cocaine, and angry l. a. insula +, ̂ (–)scenarios, two trials r. Nac/SCC +,^(–)each, same order) OFC 0

DLPFC 0Cerebellum 0

Medical College of Wisconsin

Garavan et al., fMRI Nontreatment 4’ nature video, amygdala +1998–2000 (84) (AFNI) (n = 17; followed by l. cingulate +

14 controls) 4’ cocaine/sex caudate +videos; 5” distraction parietal +task in between frontal +

(Continued)


TABLE 110.1. (Continued)

Imaging Cocaine Days ofLaboratory Technology Population Cessation Cue Description Resultsa

Brookhaven LaboratoriesWang et al., PET, FDG Nontreatment Average, 7 ± 9 30’ family genogram OFC +

1999 (85) (ROI) (n = 13) (self-report) interview 30’ “cocaine l. insula +preparation ritual” r. insulainterview with para- cerebellum +phernalia (separate days DLPFC 0for drug/nondrug cues) a. cingulate

Harvard–MassachussetsGeneral

Breiter et al., fMRI, BOLD Nontreatment 18 h minimum saline infusion in an fMRI NAc/SCC +1997–1998 (64) (n = 4, retest magnet where cocaine had r. insula +

subgroup) previously been administered OFC +

aResults legend: +, activated; 0, no difference; —, deactivated; ^, positive correlation with cocaine craving; ^(–), negative correlation withcocaine craving. When controls were studied, the summary reflects significant difference between groups for the drug cue vs. nondrug cueconditions. When no controls were studied, the effects are only for drug cue vs. nondrug cue condition in cocaine users. Because of spaceconstraints, results are summarized and “no difference” regions are presented only as relevant to discussion in the text. Please refer to the originalarticles for a complete listing of neuroanatomic regions studied.Abbreviations: SPECT, single-photon emission computed tomography; 123I-IBZM, iodobenzamide, a D2-receptor ligand; PET, positron emissiontomography; ROI, region of interest analysis; OFC, orbitofrontal cortex; DLPFC, dorsolateral prefrontal cortax; VLPFC, ventrolateral prefrontalcortex; fMRI, functional magnetic resonance image; BOLD, blood oxygen level-dependent technique; SPM, statistical parametric mappingtechnique; FDG, 12F-fluorodeoxyglucose; NAc/SCC, nucleus accumbens/subcallosal cortex; l., left; r., right; a., anterior, m., medial.

to cocaine cues often occurs in a particular subset of regionsactivated by cocaine itself (summarized above). Because ofthe number of studies, the findings reviewed below are or-ganized according to the anatomic structures that have fre-quently been activated during cue-induced craving.

Limbic-Related Structures

Several of the structures activated during cue-induced crav-ing are parts of an interconnected rostral limbic system,important in motivation and affective experience. Devinskyet al. (75) describe the rostral limbic system as including‘‘the amygdala and septum, and orbitofrontal, anterior in-sula, and anterior cingulate cortices, the ventral striatum in-cluding the nucleus accumbens, and several brainstem motornuclei including the periaqueductal gray.’’ We also includefindings for the hippocampus, part of the caudal limbicsystem in Devinsky’s organizational scheme. Several of theselimbic structures (e.g., amygdala, anterior cingulate and or-bitofrontal cortex) comprise subdivisions of functional sig-nificance, but most in vivo neuroimaging studies refer tothe structures in their entirety because of limited spatialresolution.

AmygdalaThe amygdala, located in the medial aspect of the temporallobe, is interconnected with the other rostral limbic regionsand with the hippocampus. The amygdala is critical in signallearning for biologically significant (pleasant or unpleasant)events (76,77) and has been activated by cue-induced crav-

ing in virtually all the studies able to visualize it. In the firstPET study of craving, predicted increases in amygdala rCBF(measured with 15O-water as the perfusion tracer) werefound in cocaine patients viewing videos that induced crav-ing (averaging 4.5 on a scale of 0 to 9 for ‘‘craving or desirefor cocaine’’) (47,78,79) versus nature videos. This effectwas not evident in controls without a cocaine history. Inter-estingly, baseline rCBF in the amygdala of cocaine userstended to be lower than in controls, such that increasedrCBF in response to the cocaine cues did not exceed theamygdala rCBF in control subjects under the same condi-tions. Activation of the amygdala in this study was initiallydocumented by a region-of-interest (ROI) analysis and hasrecently been confirmed by statistical parametric mapping(SPM) of the group data (Fig. 110.1, top image). The amyg-dala activation to cocaine video cues documented withfMRI (Fig. 110.2) has recently been replicated in an ongo-ing study (Listerud et al., unpublished data). In other PETstudies, Schweitzer et al. (81) and Kilts et al. (82) foundamygdala activation during memory of cocaine-inducedcraving (‘‘How strong was the urge to use cocaine, on ascale of 0 to 10?’’) induced by guided drug imagery, andGrant et al. (83) found a positive correlation of medial tem-poral lobe activation with video- and paraphernalia-inducedcraving (‘‘. . . craving or urge to use cocaine, on a scale of0 to 10’’). A recent fMRI study by Garavan et al. (84)found differential activation of the temporal pole, a regionsurrounding the ventral amygdala, in response to cocaineversus nature video. The PET camera used in the recent18F-fluorodeoxyglucose study by Wang et al. (85) did not


FIGURE 110.1. Amygdala and anterior cingulate activations dur-ing craving triggered by cocaine cues. Statistical parametric map(SPM 96) shows differential activation of brain regions by a co-caine video and a nondrug (nature) video; p �.05, corrected. Seecolor version of figure.

FIGURE 110.2. Functional magnetic resonance imag-ing of cocaine versus nature video. Individual differ-ence maps show amygdala and anterior cingulate acti-vation in three pilot cocaine patients. See color versionof figure.

permit adequate resolution of the amygdala, and susceptibil-ity artifact in the region of the amygdala prevented reliableimaging in an earlier fMRI study by Maas et al. (86).

Anterior CingulateThe anterior cingulate is located in the dorsomedial prefron-tal cortex and is interconnected with other rostral limbicstructures, including the amygdala and nucleus accumbens.Multiple roles of the anterior cingulate include selective at-tention and emotional reactivity to significant stimuli (75,87). In parallel with the amygdala findings, the anteriorcingulate was differentially activated during video-inducedcocaine craving in our initial PET study with 15O bolus (byROI analysis and later confirmation by SPM analysis of thegroup data) (Fig. 110.1, bottom image). As with the amyg-dala effect, the cue-induced rCBF increase in anterior cingu-late was from a resting baseline that was hypoactive relativeto the baseline of controls. Unpublished fMRI data (List-erud et al.) support the anterior cingulate activation in re-sponse to cocaine versus nature cues. Studies by Maas et al.(86), Kilts et al. (82), andGaravan et al. (84) have confirmedanterior cingulate activation during cocaine cues. Thesestudies used fMRI (84,86) and PET with 15O bolus (82),imaging techniques that provide good temporal resolution.Studies by Grant et al. (83) and Wang et al. (85) did notdetect increased anterior cingulate activation during cocainecues. They both used PET with 18F-fluorodeoxyglucose;because of its low temporal resolution, this technique maybe insensitive to a relatively brief or nonhomogeneous acti-vation of the anterior cingulate.


Nucleus Accumbens/Subcallosal Cortex; StriatumThe nucleus accumbens is located in the forebrain, in theventral part of the striatum. It is a prominent terminal re-gion for DA cells projecting from the ventral tegmental area,and much animal research points to this mesolim-bic–nucleus accumbens pathway as a critical substrate forthe reinforcing effects of natural rewards (88), cocaine, andother drugs of abuse (89). In humans, the size of the nucleusaccumbens is about 5 mm. Thus, the nucleus accumbensfalls at (or often below) the threshold of reliable detectionfor many early PET cameras, but it often can be localizedwith the precise anatomic co-registration of fMRI. The1997 fMRI study of cocaine administration by Breiter etal. (64) contained an ‘‘unintentional’’ cue paradigm (Table110.1). Subjects given a (double-blinded) infusion of salinesolution in the fMRI magnet showed clear activation ofthe nucleus accumbens if they had previously received aninfusion of cocaine in this novel environment. The nucleusaccumbens signal to the infusion environment had the same‘‘early-onset, sustained’’ pattern as the signals to actual co-caine administration. This striking finding suggests that adruglike response to cocaine cues can be established with asingle trial.

Other striatal findings in cue paradigms are mixed. Inthe study performed with PET and 15O bolus, the overallresponse of the basal ganglia to cocaine versus nature videoswas an unpredicted decrease in rCBF, but the larger dorsalstriatum (which receives primary projections from the sub-stantia nigra) was not parsed from the smaller ventral (nu-cleus accumbens) portion, which receives projections fromthe ventral tegmental area (47,78,79). Garavan et al. (84)found activation of the dorsal striatum (caudate) with fMRI,but did not report on nucleus accumbens. Kilts et al. (82)found nucleus accumbens activation in response to guidedcocaine imagery, but the activation was inversely correlatedwith craving self-reports.

Orbitofrontal CortexThe orbitofrontal cortex is located in the ventromedial re-gion of the frontal lobes. The orbitofrontal cortex is richlyinterconnected with DA-related regions involved in rewardand stimulus–reward learning (90). Three studies (64,83,85) found orbitofrontal cortex activation in response tocues; two did not (47,82). The remaining two studies, inwhich fMRI was used, did not report on the orbitofrontalcortex response (84,86) (ventral orbital regions are oftendifficult to image with fMRI because of artifact introducedby air in the sinus cavities). In the three studies finding anorbitofrontal cortex response to cues, the subjects were inearly cessation (ranging from 18 hours to 7 days); in thetwo studies finding no orbitofrontal cortex activation tocues, the patients had been abstinent for longer periods. Inthe earlier ‘‘cessation’’ section of this chapter, we mentionedthat the literature suggests that orbitofrontal cortex hyperac-tivity in early cocaine cessation is correlated with self-

reported craving, whereas (later) orbitofrontal cortex hypo-activity is unrelated to craving. One interpretation that inte-grates the earlier observations with those in the explicit cueparadigms is that orbitofrontal cortex hyperactivity is associ-ated with enhanced responsivity to cues, whether naturallyoccurring or presented by a laboratory experiment. Orbito-frontal cortex hypoactivity, on the other hand, clearly doesnot prevent cue-induced craving and may represent a differ-ent vulnerability (see summary below).

InsulaThe insular cortex is located interior to the lateral sulcus.It is interconnected with several other regions activated bycocaine cues, including the amygdala and the cingulate andorbitofrontal cortex; it also reflects input from the viscera(autonomic nervous system) and sensory systems. Three lab-oratories have reported activation of the insula in responseto cocaine-related cues, but the effects vary. Wang et al.(85) reported activation of the left insula but found a corre-lation of craving only with the right insula. Breiter et al.(64) reported activation of the right insula in response tothe cocaine infusion environment; no correlation with crav-ing was reported for the small subgroup in this experimentalcondition. Kilts et al. (82) reported activation of the leftinsula (in response to guided imagery), but a negative corre-lation with craving. Given the disparate findings, additionalstudies will be needed to sort out the nature and directionof cue effects in the insula.

HippocampusThe hippocampus is located in the medial temporal lobe,posterior to the amygdala. It was not differentially activatedby the cocaine videos in our PET study with 15O bolus, andactivation has not been reported by the other investigations.This suggests that cue paradigms generally do not makedemands on explicit, declarative memory and factual recall,functions closely associated with the hippocampus (91).This is in contrast to the common finding of amygdalaactivation across several cue studies. Although proximal tothe hippocampus and interconnected to it, the amygdala isnot activated by explicit memory demands; rather, it sup-ports functions of implicit, emotional memory (92).

Other Structures

Dorsolateral Prefrontal CortexThe dorsolateral prefrontal cortex, best known for its rolein working memory, was not differentially activated by theuninterrupted, narrative cocaine videos in our PET study(47), although craving was robust. Similarly, it was not acti-vated by the paradigm of Kilts et al. (82) with guided im-agery, by that of Wang et al. (85) with cocaine theme inter-views, or by that of Garavan et al. (84) with anuninterrupted, unrepeated 4-minute video clip.

The paradigm of Maas et al. (86) did activate the dorso-


lateral prefrontal cortex; it featured brief alternating expo-sures to a nondrug and a cocaine video, modified from thetapes of Childress et al. (the faces were blurred to protectthe patients’ identities). These alternating conditions mayhave engaged working memory in the cocaine subjects be-cause the same cocaine users reappeared in an ongoing drugscenario that was interrupted by the nondrug video seg-ments. Controls are generally less engaged by cocaine stim-uli and therefore would also be expected to show less engage-ment of working memory. A similar explanation mayaccount for activation of the dorsolateral prefrontal cortexin the paradigm of Grant et al. (83), which featured severalrepetitions of the same brief cocaine video clip during theperiod of 18F-fluorodeoxyglucose uptake. In an ongoingfMRI study (Listerud et al., unpublished data), activationsby uninterrupted versus alternating cocaine videos are beingcompared within the same cocaine patients. This study willdirectly test the hypothesis that dorsolateral prefrontal cor-tex activation in cocaine users is related to the mode ofstimulus presentation rather than to cue-induced cravingper se.

CerebellumThe cerebellum, important in motor coordination and re-tention of simple motor schemas, was not activated in ourPET study with 15O bolus in which videos were used toinduce craving for cocaine (47). Similarly, it was not acti-vated by the cues of Maas et al. (86), Kilts et al. (82), orGaravan et al. (84). The cerebellum was differentially acti-vated in the study of Wang et al. (85), which featured acocaine theme interview and handling of paraphernalia dur-ing the period of 18F-fluorodeoxyglucose uptake. The(highly ritualized and overlearned) handling of cocaine para-phernalia may have triggered motor memories and schemas,a cerebellar function. In support of this notion, the studyof Grant et al. (83) also featured handling of paraphernaliaas part of the stimulus complex during the period of 18F-fluorodeoxyglucose uptake, and a correlation was found be-tween craving and cerebellar activation. This correlationwould occur if handling of paraphernalia acted both as apotent conditioned cue for drug craving and as a trigger formotor memories/highly practiced motor schemas related tococaine preparation.

Sensory/Association CortexIn addition to the regions discussed, one imaging studyhas shown differential activation of visual association areas(peristriate) by cocaine cues (83), and two studies haveshown differential activation of the inferior parietal lobe(83,84), which is sometimes implicated in working mem-ory. As additional neuroimaging studies accrue, it will beeasier to determine whether less common activations suchas these reflect a feature of the paradigms used or a featureof the target state.

Summary

The involvement of seven laboratories in neuroimaging cue-induced craving has generated a useful database for drawingpreliminary conclusions about the substrates of the stateand, importantly, for generating new hypotheses that willhelp to refine the emerging picture. Despite the variabilityin imaging techniques, analysis techniques, the abstinence/treatment status of the subjects, and the varied methodsused to induce cocaine desire, several convergent findingsfor regions of activation have been obtained. The most com-monly activated regions during cocaine cues, across the lab-oratories, were the amygdala and anterior cingulate. Twostudies that parsed the ventral (nucleus accumbens) fromthe dorsal striatum showed activation by cocaine cues inthe ventral region; the fMRI signal from the nucleus accum-bens for a ‘‘cocaine-associated environment’’ (the fMRImagnet!) was strikingly similar to that for cocaine itself.The insula has been activated in at least three cue studies,although the correlations with craving vary in direction.The orbitofrontal cortex was also activated by cues in the(three) studies of cocaine subjects in recent cessation.

The hippocampus was not regularly activated by cocainecues, which suggests that the cue-induced state does notdepend on declarative memory/factual recall. The dorsolat-eral prefrontal cortex was not activated in most of the cuestudies but was activated by cocaine cues that were intermit-tent or repeated; this activation may be relatively indepen-dent of any direct connection to cue-induced craving. Thecerebellum was not activated in most cue studies, althoughit was activated in two paradigms in which paraphernaliahandling could have triggered motor memory schemas.

The brain regions activated by cocaine cue paradigms(amygdala, anterior cingulate, nucleus accumbens, insula)do substantially overlap those activated by cocaine itself inthe fMRI study of Breiter et al. (64). This is consistent withthe druglike phenomena (autonomic arousal, mild eu-phoria, sense of a cocaine ‘‘taste’’ or ‘‘smell’’) that oftenaccompany cue-induced craving. Most cue paradigms, eventhose in which fMRI is used, have not yet described thetemporal pattern of the signals from these regions duringcues; such information would permit a more detailed com-parison of cue effects with the multiple effects of cocaine(e.g., ‘‘craving’’ vs. ‘‘rush’’ substrates). No studies have yetexamined the ventral tegmental area or basal forebrain inresponse to cues.

The orbitofrontal cortex deserves special mention in thediscussion of craving and drug motivation. Orbitofrontaldysfunction in other disorders has been associated with diffi-culties in modulating rewarded or punished behavior (e.g.,reversing or switching behavior in response to a change incontingencies) (93), impaired somatic/emotional responsein anticipation of the consequences of a decision (‘‘futureinsensitivity’’) (94), and perseverative, compulsive behaviors(67). Clinically, some of these same difficulties have been


noted in substance abusers, which raises the possibility ofa core deficit in some patients (95). Long-term users ofamphetamine are impaired on a decision-making task thatplaces demands on ventromedial prefrontal (orbital) func-tion (96), and long-term administration of stimulantsclearly erodes orbitostriatal inhibitory function in primates(93). Whether such orbitofrontal cortex deficits predatedrug use in humans, predispose to it, or are a consequenceof it, the news for long-term cocaine users is not good;they may be at a particular disadvantage in managing theircraving for the drug. Interestingly, the reviewed studies linkincreases in orbitofrontal cortex activity to craving in allthree paradigms: (early) cessation, stimulant administration,and response to cues (during early cessation from cocaine).A future challenge is to understand how such activations ina potentially dysfunctional orbitofrontal cortex differ (ornot) from normal activation of the orbitofrontal cortex dur-ing advantageous decision making (97).

DISCUSSION

Theoretic Implications

The neuroimaging of craving states is at a very early stage,and most findings should be replicated before they are takenas either a confirmation or a challenge to theories of addic-tion and drug motivation. With this caveat kept in mind,some findings from the paradigms reviewed may have impli-cations for current theories.

From the cessation paradigms, the preliminary imagingdata provide little support for the proposition that craving(drug desire) during cessation arises primarily from a stateof prolonged DA depletion or deficiency. Although the DAsystems of cocaine patients do differ from those of controls,craving did not show a good relationship to these changesbeyond the first week of cessation.

The drug administration paradigms show that DA-relatedregions are activated during stimulant-induced craving, butthe temporal pattern of activation fits neither a simple prim-ing (‘‘substrates for craving are the same as the substratesfor high’’) nor a simple opponent process (‘‘substrates forcraving are the opposite of the substrates for high’’) view.Studies of craving during methylphenidate administrationindicate the possible contribution of non-DA (in additionto DA) systems.

The cue paradigms suggest that the regions activated dur-ing cue-induced craving often overlap the regions activatedby cocaine itself (i.e., the substrate is substantially ‘‘drug-like’’). Of course, these early findings do not preclude‘‘drug-opposite’’ brain responses or conditioning in re-sponse to cocaine cues. The cue data simply suggest thatacross several laboratory environments (as in patients’ de-scriptions from the natural environment), cocaine cravingwith a ‘‘druglike’’ substrate is likely to be evoked.

Interestingly, the current neuroimaging paradigms offer

little support for the notion of a ‘‘sensitized’’ substrate,sometimes proposed as a potential mechanism for stimulantdrug craving/incentive motivation. Beyond the first weekof cessation, cocaine patients often exhibit resting hypoac-tivity in limbic and frontal regions in comparison with con-trols. Exposure to cocaine cues or to a stimulant can producea significant activation in these affected brain regions, butthe absolute level of brain response is often no greater thanin controls, and may even be less. Of course, the appropriatetest for sensitization would require comparing the currentresponses of cocaine patients with their own initial responsesto the drug. This design is unfortunately not feasible. Itdoes, however, highlight a limitation of all our neuroimag-ing studies in cocaine patients; we usually do not know theirbaseline before addiction on the brain variables of interest.The discovery of Volkow et al. (50) of striking baselinedifferences in the availability of D2 receptors in controls(based on liking vs. not liking an initial dose of methylphen-idate) is fair warning that the brains of cocaine users whoproceed to addiction may differ from those of (some) con-trols, even before cocaine use begins.

Treatment Implications

Imaging data from the stimulant administration and cueparadigms suggest that craving is often associated with rela-tive increases in the activity of the same brain DA systemsthat may otherwise be hypoactive in cocaine users. Findingan ‘‘anticraving’’ agent that can modulate the periodic in-creases in DA in response to cues or drug primes, withoutworsening symptoms that may be related to chronic hypoac-tivity (e.g., depressed mood, low levels of energy), poses aparticular challenge. Classic DA antagonists (typified by theolder antipsychotics) are poor candidates for this modula-tory role because they could worsen symptoms related tolow levels of DA; they also carry a significant risk for extra-pyramidal side effects and tardive dyskinesia.

Partial DA agonists may offer an appealing solution inthe future (2,98,99). These compounds act as agonistsunder conditions of low DA tone (as may occur in cessa-tion), but as antagonists when the DA concentration in-creases (as in response to cues or drug primes). The ‘‘chame-leon-like’’ nature of partial agonists may possibly offer thecocaine patient a moment-to-moment regulation of the DAsystem. Unfortunately, for now, no partial agonists (D1,D2, or D3) are approved for humans, although considerableanimal research has been done and preliminary safety trialsare under way.

Another promising category of DA modulators are the�-aminobutyric acid type B (GABAB) agonists (100–103).These compounds may gently modulate the DA system byreducing ventral tegmental area cell firing, thereby reducingDA release in terminal regions. Roberts and colleagues (100,101) were the first to demonstrate the blunting of cocainemotivation by the GABAB agonist baclofen (Fig. 110.3). In


FIGURE 110.3. Baclofen, a �-aminobu-tyric acid type B agonist, may blunt lim-bic activation and craving in response tococaine cues. See color version of figure.

subsequent cocaine-related studies by Dewey et al. (104),the GABA transaminase inhibitor �-vinyl-GABA alsoshowed promise; its cocaine-related effects are reversible bya GABAB antagonist. Unpublished preliminary data fromChildress et al., testing the ability of baclofen to blunt bothsubjective and brain responses to cocaine cues by measuringrCBF with PET and 15O bolus, suggest that baclofen, al-though it has a relatively short half-life, may indeed conferprotection against cue-induced craving and the accompany-ing limbic activation. These data are important because thecraving/imaging paradigm is being used to test an ‘‘anticrav-ing’’ medication.

Future Directions

The neuroimaging studies of cocaine craving reviewed inthis chapter will soon be viewed as the ‘‘early’’ stage of ourunderstanding; the imaging field is growing rapidly, as isthe sophistication of the tools and their users. Advancesin spatial and temporal resolution of imaging devices, andadvances in image analysis, will allow the formulation ofmore precise hypotheses regarding craving substrates. Asshown in this review, the future answers are likely to befound within temporal as well as regional patterns of activa-tion. Although DA has played a strong role in shaping theearly neurochemical hypotheses, interacting neurotransmit-ters and neuromodulators will soon be tested as the criticalligands become available. Until then, the combination ofpharmacologic probes with neuroanatomic imaging mayoffer a powerful alternative.

Designs of increased rigor, with attention given to homo-geneity of samples (e.g., number of days of cessation, nico-tine status, treatment status, urine toxicology status, geneticstatus, drug use history) and careful characterization of con-trols, will enhance the replicability of findings across labora-tories. Asking for more than one subjective response or

‘‘craving item,’’ and asking for these at the optimal timefor the paradigm, will ensure a clear test of the relationshipbetween brain activity and subjective state. Characterizingbrain activity during other, nondrug states of arousal (e.g.,anger, anxiety) will help to determine the specificity of thesignature of the brain for cue-induced craving. This is im-portant because the brain structures activated in cue-induced cocaine craving are not ‘‘reserved’’ for this state;rather, they participate in many other states that are notrelated to cocaine. In this regard, measurement of the brainresponse during other, nondrug appetitive states (e.g., sexualdesire) in subjects who do not use drugs may provide a‘‘positive control’’ for cocaine craving, which is so oftendescribed in sexual terms. Imaging of the craving states forheroin, nicotine, and other drugs of abuse will also provideinformative comparisons; these studies have already begun(105).

Although the path ahead is clearly challenging, findingthe brain substrates of desire, both for drugs of abuse andfor natural rewards, is now a matter of time and effort; thetools are increasingly available. Only a decade ago, and forall of prior human history, brain activity during subjectivestates was largely a matter of inference. Now, and in thefuture, these states can be the subject of direct measure.This represents a dramatic paradigm shift, one that enablesstates such as ‘‘desire’’ and ‘‘craving’’ to be the subject ofrigorous scientific research. This research is a critical prereq-uisite to the rational, and vastly improved, treatment ofdisorders of desire (i.e., the addictions).

ACKNOWLEDGMENTS

We thank V. Mikhalovsky, A. Meyer, and G. Robinson forcyclotron operations and preparation of 15O; S. Riggins, K.Kilroy, D. Dines-Meehan, and C. Herman for PET opera-


tions; W. McElgin for physics consultation; J. Fitzgerald,S. McDonald, J. D. Gray, and A. Fornash for research assis-tance; and M. Bromwell for editing. This research was sup-ported by research grants NIDA RO1 10241 to Dr.Childress and Core of NIDA P-60 Center to Dr. O’Brien,and by the Medical Research Division of the PhiladelphiaDepartment of Veterans Affairs Medical Center.

REFERENCES

1. Kosten TT, McCance E. A review of pharmacotherapies forsubstance abuse. Am J Addictions 1996;5:58–65.

2. Childress AR, O’Brien CP. Dopamine partial agonists may ad-dress the duality of cocaine craving. Trends Pharmacol Sci 2000.

3. Gawin FH, Kleber HD. Abstinence symptomatology and psy-chiatric diagnosis in cocaine abusers. Arch Gen Psychiatry 1986;43:107–113.

4. Jaffe JJ, et al. Cocaine-induced cocaine craving. Psychopharma-cology 1989;97:59–64.

5. Childress AR, et al. Conditioned craving and arousal in cocaineaddiction: a preliminary report. NIDA Res Monogr 1987, 1988;(ADM) 88-1564:74–80.

6. Ehrman RN, et al. Conditioned responses to cocaine-relatedstimuli in cocaine abuse patients. Psychopharmacology 1992;107:523–529.

7. Parsons LH, Smith AD, Justice JB. Basal extracellular dopamineis decreased in the rat nucleus accumbens during abstinencefrom chronic cocaine. Synapse 1991;9:60–65.

8. Markou A, Koob GF. Post-cocaine anhedonia. An animalmodel of cocaine withdrawal. Neuropsychopharmacology 1991;4:17–26.

9. Dackis CA, Gold MS. New concepts in cocaine addiction. Thedopamine depletion hypothesis. Neurosci Biobehav Rev 1985;9:469–477.

10. Sherer MA, Kumor KM, Jaffe J. Effects of intravenous cocaineare partially attenuated by haloperidol. Psychiatry Res 1989;27:117–125.

11. Childress AR, et al. Brain correlates of induced cocaine craving.Soc Neurosci Abstr 1993; (abst No. 762.8).

12. Berger SP, et al. Haloperidol antagonism of cue-elicited craving.Lancet 1996;347:504–508.

13. Rocha BA, et al. Cocaine self-administration in dopamine trans-porter knockout mice. Nat Neurosci 1998;1:132–137.

14. Sora I, et al. Cocaine reward models: conditioned place prefer-ence can be established in dopamine- and in serotonin-knockoutmice. Proc Natl Acad Sci U S A 1998;95:7699–7704.

15. Glowa J, et al. Phentermine/fenfluramine decreases cocaine self-administration in rhesus monkeys. Neuroreport 1997;8:1347–1351.

16. Satel SL, et al. Tryptophan depletion and attenuation of cue-induced craving for cocaine. Am J Psychiatry 1995;152:778–783.

17. Kalivas PW, Duffy P. Repeated cocaine administration altersextracellular glutamate in the ventral tegmental area. J Neuro-chem 1998;70:1497–1502.

18. Piazza PV, et al. Stress, glucocorticoids, and mesencephalic do-paminergic neurons: a pathophysiological chain determiningvulnerability to psychostimulant abuse. In: Neurotoxicity andneuropathology associated with cocaine abuse. NIH publicationNo. 96-4019, 1994.

19. Goeders N. A neuroendocrine role in cocaine reinforcement.Psychoneuroendocrinology 1997;22:237–259.

20. Shaham Y, et al. CP-154,526, a selective, non-peptide antago-

nist of the corticotropin-releasing factor 1 receptors attenuatesstress-induced relapse to drug seeking in cocaine- and heroin-trained rats. Psychopharmacology 1998;137:184–190.

21. Spealman RD, Bergman J. Modulation of the discriminativestimulus effects of cocaine by mu and kappa opioids. J Pharma-col Exp Ther 1993;261:607–615.

22. Shippenberg TS. Sensitization to the behavioral effects of co-caine: modulation by dynorphin and kappa-opioid receptor ago-nists. Pharmacol Biochem Behav 1997;57:449–455.

23. Roberts DCS, Corcoran ME, Fibiger HC. On the role of as-cending catecholamine systems in the self-administration of co-caine. Pharmacol Biochem Behavior 1977;6:615–620.

24. Wise RA, Bozarth MA. Brain substrates for reinforcement anddrug-self-administration. Prog Neuropsychopharmacol 1981;5:467–474.

25. Koob GF, Bloom FE. Cellular and molecular mechanisms ofdrug dependence. Science 1988;242:715–723.

26. Fowler JS, et al. Positron emission studies of dopamine-enhanc-ing drugs. J Clin Pharmacol 1999. 13S–16S.

27. Zubieta J-K, et al. Increased mu opioid receptor binding de-tected by PET in cocaine-dependent men is associated withcraving. Nat Med 1996;2:1225–1229.

28. Ritz MC, Lamb RJ, et al. Cocaine receptors on dopamine trans-porters are related to self-administration of cocaine. Science1987;237:1219–1223.

29. Volkow ND, et al. Relationship between subjective effects ofcocaine and dopamine transporter occupancy. Nature 1997;386:827–830.

30. Baxter LR, Schwartz JM, Phelps M. Localization of neurochem-ical effects of cocaine and other stimulants in the human brain.J Clin Psychiatry 1988;49:23–26.

31. Volkow ND, et al. Cocaine uptake is decreased in the brain ofdetoxified cocaine abusers. Neuropsychopharmacology 1996;14:159–168.

32. Volkow ND, et al. Effects of chronic abuse on postsynapticdopamine receptors. Am J Psychiatry 1990;147:719–724.

33. VolkowND, et al. Decreased dopamine D2 receptor availabilityis associated with reduced frontal metabolism in cocaine abusers.Synapse 1993;14:169–177.

34. Unterwald EM, et al. Chronic cocaine alters brain mu opioidreceptors. Brain Res 1992;584:314–318.

35. Hurd YL, HarkenhamM. Molecular alterations in the neostria-tum of human cocaine addicts. Synapse 1993;13:357–369.

36. Hammer RJ. Chronic cocaine alters opiate receptor binding incritical brain reward regions. Synapse 1989;8:55–60.

37. Little KY, et al. Cocaine use increases (3-H) WIN 35428 bind-ing sites in human striatum. Brain Res 1993;628:17–25.

38. Staley JK, et al. High-affinity cocaine recognition sites on thedopamine transporter are elevated in fatal cocaine overdose vic-tims. J Pharmacol Exp Ther 1994;271:1678–1685.

39. Malison RT, et al. Elevated striatal dopamine transporters dur-ing acute cocaine abstinence as measured by I-123 B-CITSPECT. Am J Psychiatry 1998;155:832–834.

40. Hitri A, et al. Fewer dopamine transporter receptors in theprefrontal cortex of cocaine users. Am J Psychiatry 1994;151:1074–1076.

41. Wilson JM, et al. Striatal dopamine transporter, and vesicularmonoamine transporter, in chronic cocaine users. Ann Neurol1996;40:428–439.

42. Markou A, Koob GF. Bromocriptine reverses the elevation inintracranial self-stimulation thresholds observed in a rat modelof cocaine withdrawal. Neuropsychopharmacology 1992;7:213–224.

43. Hammer RP, et al. Withdrawal following cocaine self-adminis-tration decreases regional cerebral metabolic rate in critical brainreward regions. Synapse 1993;14:73–80.


44. Trulson ME, Ulissey JJ. Chronic cocaine administration de-creases dopamine synthesis rate and increases (H) spiroperidolbinding in rat brain. 1987.

45. Volkow ND, et al. Changes in brain glucose metabolism incocaine dependence and withdrawal. Am J Psychiatry 1991;148:621–626.

46. Volkow ND, et al. Long-term frontal brain metabolic changesin cocaine abusers. Synapse 1992;11:184–190.

47. Childress AR, et al. Limbic activation during cue-induced co-caine craving. Am J Psychiatry 1999;156:11–18.

48. Childress AR, et al. Unpublished data.49. Gawin F, Riordan C, Kleber HD. Methylphenidate treatment

of cocaine abusers without attention deficit disorder: a negativereport. Am J Drug Alcohol Abuse 1985;11:193–197.

50. Volkow ND, et al. Prediction of reinforcing responses to psy-chostimulants in humans by brain dopamine D2 receptor levels.Am J Psychiatry 1999;156:1440–1443.

51. Volkow ND, et al. Decreased striatal dopaminergic responsive-ness in detoxified cocaine abusers. Nature 1997;386:830–833.

52. Eikelboom R, Stewart J. Conditioning of drug-induced physio-logical responses. Psychol Rev 1982;89:507–528.

53. Koob GF. Drug addiction: the yin and yang of hedonic homeo-stasis. Neuron 1996;16:893–896.

54. Gardener EL. Brain reward mechanisms. In: Lowinson J, RuizP, Millman R, eds. Comprehensive textbook of substance abuse.Baltimore: Williams & Wilkins, 1993:70–99.

55. Roberts DCS, Ranaldi R. Effect of dopaminergic drugs on co-caine reinforcement. Clin Neuropharmacol 1995;18:S84–S95.

56. Wise RA. Neurobiology of addiction. Curr Opin Neurobiol1996;6:243–251.

57. Volkow ND, et al. Blockade of striatal dopamine transportersby intravenous methylphenidate is not sufficient to induce self-reports of ‘‘high.’’ J Pharmacol Exp Ther 1999;288:14–20.

58. Pearlson GD, et al. Correlation of acute cocaine-inducedchanges in local cerebral blood flow with subjective effects. AmJ Psychiatry 1993;150:495–497.

59. Volkow ND, et al. Is methylphenidate like cocaine? Studies ontheir pharmacokinetics and distribution in the human brain.Arch Gen Psychiatry 1995;52:456–463.

60. Malison RT, et al. Euphorigenic doses of cocaine reduce (123I)�-CIT SPECT measures of dopamine transporter availability inhuman cocaine addicts. Psychopharmacology 1995;122:358–362.

61. Schlaepfer TE, et al. PET study of competition between intrave-nous cocaine and C-11 raclopride at dopamine receptors inhuman subjects. Arch Gen Psychiatry 1997;154:1209–1213.

62. London ED, et al. Cocaine-induced reduction of glucose utiliza-tion in human brain. Arch Gen Psychiatry 1990;47:567–574.

63. Kaufman MJ, et al. Cocaine-induced cerebral vasoconstrictiondetected in humans with magnetic resonance angiography.JAMA 1998;279:376–380.

64. Breiter HC, et al. Acute effects of cocaine on human brainactivity and emotion. Neuron 1997;19:591–611.

65. Deleted in press.66. Volkow ND, et al. Association of methylphenidate-induced

craving with changes in right striato-orbitofrontal metabolismin cocaine abusers: implications in addiction. Am J Psychiatry1999;156:19–26.

67. Volkow ND, Fowler JS. Addiction, a disease of compulsionand drive: involvement of the orbitofrontal cortex. Cereb Cortex2000;10:318–325.

68. Volkow ND, et al. Relationship between psychostimulant-in-duced high and dopamine transporter occupancy. Proc NatlAcad Sci U S A 1996;93:1–5.

69. Childress AR, et al. Classically conditioned factors in drug de-pendence. In: Lowinson J, Ruiz P, Millman R, eds. Comprehen-

sive textbook of substance abuse. Baltimore: Williams & Wilkins,1993:56–69.

70. Weiss F, et al. Control of cocaine-seeking behavior by drug-associated stimuli in rats: effects on recovery of extinguishedoperant-responding and extracellular dopamine levels in amyg-dala and nucleus accumbens. Proc Natl Acad Sci 2000;97:4321–4326.

71. Franklin TR, Druhan JP. Expression of Fos-related antigens inthe nucleus accumbens and associated regions following expo-sure to a cocaine-paired environment. Eur J Neurosci 2000;6:2097–2106.

72. Brown EE, RobertsonGS, FibigerHC. Evidence for conditionalneuronal activation following exposure to a cocaine-paired envi-ronment: role of forebrain limbic structures. J Neurosci 1992;12:4112–4121.

73. Childress AR, et al. Brain imaging of withdrawal and cravingstates in cocaine dependence. In: Proceedings of the 54th annualmeeting of the Committee on Problems of Drug Dependence, Key-stone, Colorado, 1992.

74. Childress AR, et al. Unpublished data.75. Devinsky O, Morrell MJ, Vogt BA. Contributions of anterior

cingulate cortex to behaviour. Brain 1995;118:279–306.76. Everitt BJ, Robbins TW. Amygdala-ventral striatal interactions

and reward-related processes. In: Aggleton JP, ed. The amygdala.New York: Wiley-Liss, 1992:401–430.

77. LeDoux JE. Emotion and the amygdala. In: Aggleton JP, ed.The amygdala. New York: Wiley-Liss, 1992:339–351.

78. Childress AR. Brain imaging during drug-craving states. In:Symposium on brain imaging in substance abuse, 56th annualmeeting of the Committee on Problems of Drug Dependence,West Palm Beach, FL, 1994.

79. Childress AR, et al. Regional brain blood flow during inducedcocaine craving. Soc Neurosci Abstr 1994;20:1632(abst No.666.2).

80. Listerud et al., 2000.81. Schweitzer J, et al. The neuroanatomy of drug craving in crack

cocaine addiction: a PET analysis. Soc Neurosci Abstr 1996;22:933(abst No. 365.3).

82. Kilts CD, et al. Functional anatomy of drug craving in humancocaine addiction. Arch Gen Psychiatry 2000 (in press).

83. Grant S, et al. Activation of memory circuits during cue-elicitedcraving. Proc Natl Acad Sci U S A 1996;93:12040–12045.

84. Garavan H, et al. Cue-induced cocaine craving: neuroanatomi-cal specificity for drug-users and drug stimuli. Am J Psychiatry2000 (in press).

85. Wang GJ, et al. Regional brain metabolic activation duringcraving elicited by recall of previous drug experiences. Life Sci1999;64:775–784.

86. Maas LC, et al. Functional MRI of human brain activationduring cue-induced cocaine craving. In: Proceedings of the Collegeon Problems of Drug Dependence, Nashville, TN, 1996.

87. Vogt BA, Finch DM, Olson CR. Functional heterogeneity incingulate cortex: the anterior executive and posterior evaluativeregions. Cereb Cortex 1992;2:435–443.

88. Schultz W, Dayan P, Montague PR. A neural substrate of pre-diction and reward. Science 1997;275:1593–1599.

89. Koob GF. Neural mechanisms of drug reinforcement. Ann NY Acad Sci 1992;654:171–191.

90. Rolls ET. The orbitofrontal cortex and reward. Cereb Cortex2000;10:284–294.

91. Squire L, et al. Activation of the hippocampus in normal hu-mans: a functional anatomical study of memory. Proc Natl AcadSci U S A 1992;89:1837–1841.

92. LeDoux JE. Emotional memory systems in the brain. BehavBrain Res 1993;58:69–79.


93. Jentsch JD, Taylor JR. Impulsivity resulting from frontostriataldysfunction in drug abuse: implications for the control of behav-ior by reward-related stimuli. Psychopharmacology 1999;146:373–390.

94. Bechara A, Damasio H, Damasio A. Emotion, decision-makingand the orbitofrontal cortex. Cereb Cortex 2000;10:295–307.

95. Monterosso. Unpublished data.96. Rogers RD, et al. Dissociable cognitive deficits in the decision-

making cognition of chronic amphetamine abusers, opiateabusers, patients with focal damage to the prefrontal cortex, andtryptophan-depleted normal volunteers: evidence for monoami-nergic mechanisms. Neuorpsychopharmacology 1999;20:323–339.

97. Rogers RD, et al. Choosing between small, likely rewards andlarge, unlikely rewards activates inferior and orbital prefrontalcortex. J Neurosci 1999;20:9029–9038.

98. Pulvirenti L, Koob GF. Dopamine receptor agonists, partialagonists, and psychostimulant addiction. Trends Pharmacol Sci1994;15:374–379.

Neuropsychopharmacology: The Fifth Generation of Progress. Edited by Kenneth L. Davis, Dennis Charney, Joseph T. Coyle, andCharles Nemeroff. American College of Neuropsychopharmacology � 2002.

99. Pilla M, et al. Selective inhibition of cocaine-seeking behaviourby a partial D3 dopamine receptor agonist. Nature 1999;400:371–375.

100. Roberts DCS, Andrews MM, Vickers GJ. Baclofen attenuatesthe reinforcing effects of cocaine in rats. Neuropsychopharmacol-ogy 1996;15:417–423.

101. Roberts DCS, Andrews MM. Baclofen suppression of cocaineself-administration: demonstration using a discrete trials proce-dure. Psychopharmacology 1997;131:271–277.

102. Ling W, Shoptaw S, Majewska MD. Baclofen as a cocaine-anti-craving medication: a preliminary clinical study. Neuropsycho-pharmacology 1998;18:403–404.

103. Childress AR, et al. Impact of GABAergics on brain activityduring cue-induced cocaine craving. Soc Neurosci Abstr 1999;25:815(abst No. 328.14).

104. Dewey SL, et al. A novel strategy for the treatment of cocaineaddiction. Synapse 1998;30:119–129.

105. Sell LA, et al. Activation of reward circuitry in human opiateaddicts. Eur J Neurosci 1999;11:1042–1048.

neuroimaging of cocaine craving states ......110 neuroimaging of cocaine craving states: cessation,...

Documents